Shutdown handler not registered because Python interpreter is not running in the main thread
run pipeline %s
run pipeline stage %s
Running pipeline stage MKMLizer
Starting job with name sao10k-12b-party-v2-v6-mkmlizer
Waiting for job on sao10k-12b-party-v2-v6-mkmlizer to finish
sao10k-12b-party-v2-v6-mkmlizer: ╔═════════════════════════════════════════════════════════════════════╗
sao10k-12b-party-v2-v6-mkmlizer: ║ _____ __ __ ║
sao10k-12b-party-v2-v6-mkmlizer: ║ / _/ /_ ___ __/ / ___ ___ / / ║
sao10k-12b-party-v2-v6-mkmlizer: ║ / _/ / // / |/|/ / _ \/ -_) -_) / ║
sao10k-12b-party-v2-v6-mkmlizer: ║ /_//_/\_, /|__,__/_//_/\__/\__/_/ ║
sao10k-12b-party-v2-v6-mkmlizer: ║ /___/ ║
sao10k-12b-party-v2-v6-mkmlizer: ║ ║
sao10k-12b-party-v2-v6-mkmlizer: ║ Version: 0.11.12 ║
sao10k-12b-party-v2-v6-mkmlizer: ║ Copyright 2023 MK ONE TECHNOLOGIES Inc. ║
sao10k-12b-party-v2-v6-mkmlizer: ║ https://mk1.ai ║
sao10k-12b-party-v2-v6-mkmlizer: ║ ║
sao10k-12b-party-v2-v6-mkmlizer: ║ The license key for the current software has been verified as ║
sao10k-12b-party-v2-v6-mkmlizer: ║ belonging to: ║
sao10k-12b-party-v2-v6-mkmlizer: ║ ║
sao10k-12b-party-v2-v6-mkmlizer: ║ Chai Research Corp. ║
sao10k-12b-party-v2-v6-mkmlizer: ║ Account ID: 7997a29f-0ceb-4cc7-9adf-840c57b4ae6f ║
sao10k-12b-party-v2-v6-mkmlizer: ║ Expiration: 2024-10-15 23:59:59 ║
sao10k-12b-party-v2-v6-mkmlizer: ║ ║
sao10k-12b-party-v2-v6-mkmlizer: ╚═════════════════════════════════════════════════════════════════════╝
sao10k-12b-party-v2-v6-mkmlizer: Downloaded to shared memory in 29.595s
sao10k-12b-party-v2-v6-mkmlizer: quantizing model to /dev/shm/model_cache, profile:s0, folder:/tmp/tmpbfb3cz98, device:0
sao10k-12b-party-v2-v6-mkmlizer: Saving flywheel model at /dev/shm/model_cache
sao10k-12b-party-v2-v6-mkmlizer: /opt/conda/lib/python3.10/site-packages/mk1/flywheel/functional/loader.py:55: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature.
sao10k-12b-party-v2-v6-mkmlizer: tensors = torch.load(model_shard_filename, map_location=torch.device(self.device), mmap=True)
sao10k-12b-party-v2-v6-mkmlizer: quantized model in 35.601s
sao10k-12b-party-v2-v6-mkmlizer: Processed model Sao10K/12b-party-v2 in 65.196s
sao10k-12b-party-v2-v6-mkmlizer: creating bucket guanaco-mkml-models
sao10k-12b-party-v2-v6-mkmlizer: Bucket 's3://guanaco-mkml-models/' created
sao10k-12b-party-v2-v6-mkmlizer: uploading /dev/shm/model_cache to s3://guanaco-mkml-models/sao10k-12b-party-v2-v6
sao10k-12b-party-v2-v6-mkmlizer: cp /dev/shm/model_cache/config.json s3://guanaco-mkml-models/sao10k-12b-party-v2-v6/config.json
sao10k-12b-party-v2-v6-mkmlizer: cp /dev/shm/model_cache/special_tokens_map.json s3://guanaco-mkml-models/sao10k-12b-party-v2-v6/special_tokens_map.json
sao10k-12b-party-v2-v6-mkmlizer: cp /dev/shm/model_cache/tokenizer_config.json s3://guanaco-mkml-models/sao10k-12b-party-v2-v6/tokenizer_config.json
sao10k-12b-party-v2-v6-mkmlizer: cp /dev/shm/model_cache/tokenizer.json s3://guanaco-mkml-models/sao10k-12b-party-v2-v6/tokenizer.json
sao10k-12b-party-v2-v6-mkmlizer: cp /dev/shm/model_cache/flywheel_model.0.safetensors s3://guanaco-mkml-models/sao10k-12b-party-v2-v6/flywheel_model.0.safetensors
sao10k-12b-party-v2-v6-mkmlizer:
Loading 0: 0%| | 0/363 [00:00<?, ?it/s]
Loading 0: 2%|▏ | 7/363 [00:00<00:07, 47.46it/s]
Loading 0: 4%|▍ | 16/363 [00:00<00:05, 68.00it/s]
Loading 0: 7%|▋ | 25/363 [00:00<00:04, 75.51it/s]
Loading 0: 9%|▉ | 34/363 [00:00<00:04, 76.62it/s]
Loading 0: 12%|█▏ | 43/363 [00:00<00:04, 79.45it/s]
Loading 0: 14%|█▍ | 52/363 [00:00<00:03, 80.18it/s]
Loading 0: 17%|█▋ | 61/363 [00:01<00:14, 20.52it/s]
Loading 0: 19%|█▉ | 70/363 [00:01<00:10, 27.02it/s]
Loading 0: 22%|██▏ | 79/363 [00:02<00:08, 33.44it/s]
Loading 0: 24%|██▍ | 88/363 [00:02<00:06, 40.78it/s]
Loading 0: 27%|██▋ | 97/363 [00:02<00:05, 48.37it/s]
Loading 0: 29%|██▉ | 106/363 [00:02<00:04, 55.42it/s]
Loading 0: 32%|███▏ | 115/363 [00:02<00:04, 61.38it/s]
Loading 0: 34%|███▍ | 124/363 [00:02<00:03, 65.17it/s]
Loading 0: 37%|███▋ | 133/363 [00:02<00:03, 69.63it/s]
Loading 0: 39%|███▉ | 142/363 [00:03<00:10, 21.41it/s]
Loading 0: 42%|████▏ | 151/363 [00:03<00:07, 27.41it/s]
Loading 0: 44%|████▍ | 160/363 [00:04<00:05, 34.32it/s]
Loading 0: 47%|████▋ | 169/363 [00:04<00:04, 41.62it/s]
Loading 0: 49%|████▉ | 178/363 [00:04<00:03, 48.59it/s]
Loading 0: 52%|█████▏ | 187/363 [00:04<00:03, 55.22it/s]
Loading 0: 54%|█████▍ | 196/363 [00:04<00:02, 61.21it/s]
Loading 0: 56%|█████▋ | 205/363 [00:04<00:02, 64.15it/s]
Loading 0: 59%|█████▉ | 214/363 [00:04<00:02, 67.92it/s]
Loading 0: 61%|██████▏ | 223/363 [00:05<00:06, 21.61it/s]
Loading 0: 64%|██████▍ | 232/363 [00:05<00:04, 27.90it/s]
Loading 0: 66%|██████▋ | 241/363 [00:05<00:03, 34.19it/s]
Loading 0: 69%|██████▉ | 250/363 [00:06<00:02, 41.68it/s]
Loading 0: 71%|███████▏ | 259/363 [00:06<00:02, 48.94it/s]
Loading 0: 74%|███████▍ | 268/363 [00:06<00:01, 55.48it/s]
Loading 0: 76%|███████▋ | 277/363 [00:06<00:01, 61.08it/s]
Loading 0: 79%|███████▉ | 286/363 [00:06<00:01, 66.99it/s]
Loading 0: 81%|████████▏ | 295/363 [00:06<00:00, 70.99it/s]
Loading 0: 84%|████████▎ | 304/363 [00:07<00:02, 21.71it/s]
Loading 0: 86%|████████▌ | 313/363 [00:07<00:01, 27.40it/s]
Loading 0: 89%|████████▊ | 322/363 [00:07<00:01, 34.12it/s]
Loading 0: 91%|█████████ | 331/363 [00:08<00:00, 40.29it/s]
Loading 0: 94%|█████████▎| 340/363 [00:08<00:00, 47.20it/s]
Loading 0: 96%|█████████▌| 349/363 [00:08<00:00, 54.09it/s]
Loading 0: 99%|█████████▊| 358/363 [00:08<00:00, 60.25it/s]
Job sao10k-12b-party-v2-v6-mkmlizer completed after 92.44s with status: succeeded
Stopping job with name sao10k-12b-party-v2-v6-mkmlizer
Pipeline stage MKMLizer completed in 93.51s
run pipeline stage %s
Running pipeline stage MKMLTemplater
Pipeline stage MKMLTemplater completed in 0.10s
run pipeline stage %s
Running pipeline stage MKMLDeployer
Creating inference service sao10k-12b-party-v2-v6
Waiting for inference service sao10k-12b-party-v2-v6 to be ready
Inference service sao10k-12b-party-v2-v6 ready after 190.41855335235596s
Pipeline stage MKMLDeployer completed in 190.90s
run pipeline stage %s
Running pipeline stage StressChecker
Received healthy response to inference request in 2.4860177040100098s
Received healthy response to inference request in 1.8555288314819336s
Received healthy response to inference request in 1.9925792217254639s
Received healthy response to inference request in 1.892732858657837s
Received healthy response to inference request in 1.7285492420196533s
5 requests
0 failed requests
5th percentile: 1.7539451599121094
10th percentile: 1.7793410778045655
20th percentile: 1.8301329135894775
30th percentile: 1.8629696369171143
40th percentile: 1.8778512477874756
50th percentile: 1.892732858657837
60th percentile: 1.9326714038848878
70th percentile: 1.9726099491119384
80th percentile: 2.0912669181823733
90th percentile: 2.2886423110961913
95th percentile: 2.3873300075531003
99th percentile: 2.466280164718628
mean time: 1.9910815715789796
Pipeline stage StressChecker completed in 10.63s
run pipeline stage %s
Running pipeline stage TriggerMKMLProfilingPipeline
run_pipeline:run_in_cloud %s
starting trigger_guanaco_pipeline args=%s
Pipeline stage TriggerMKMLProfilingPipeline completed in 3.89s
Shutdown handler de-registered
sao10k-12b-party-v2_v6 status is now deployed due to DeploymentManager action
sao10k-12b-party-v2_v6 status is now inactive due to auto deactivation removed underperforming models